Forecasting Urban Travel Past, Present and Future. David Boyce and Huw Williams

Transcription

1 Forecasting Urban Travel Past, Present and Future David Boyce and Huw Williams

2 How did the Book come about? We first met at the Institute for Transport Studies at the University of Leeds in 1972, and compared notes on US and UK urban transportation studies. In 2003, we began to write on a comparison between model developments in US and UK. In 2007, we finalised a book outline on the history of ideas underpinning models and methods for forecasting. In 2014, the first draft was completed!

3 What were our Objectives? Write history of the development of the field in terms of influential ideas. Consider innovations in relation to existing knowledge and spirit of the times. Explore how innovation was transferred over time and space. Draw in contributions from mathematics, statistics, physics, engineering, operational research, economics, geography and psychology. Establish a literature which we hoped would be useful to the field. Hoped to make a contribution that was widely accessible to academics, professional planners, students and the interested citizen.

4 What were our Methods? Drew on our own experience in particular areas. Consulted national and international literature, read/reread major papers and reports produced over the last 60 years. Wrote to many key figures in the development of the field. Consulted many current theoreticians and practitioners about the current states of the art in particular areas. Circulated chapters widely to check their accuracy and integrity.

5 The style of the book To fulfil our accessibility objective: No mathematics in the text (Hawking s Law). Technical details and further reading in the chapter end notes.

6 Models and their uses Representations of the behavior of a system - changes over time - changes in response to policies Policies very wide ranging Generate information of interest to planners and policy makers

7 Establish themes of development and innovation Within a broadly chronological approach, we considered the following themes: Role of institutions in developing, sponsoring and promoting models Planning contexts, information requirements Nature and relevance of theory Data requirements of models and methods Solution of models Validation and performance of models Practical compromises in model development

8 Chapter 2: Emergence of the traditional approach Impetus for urban transportation studies. Role of forecasting and plan testing in the Detroit and Chicago areas. Major advances in zone-based land use, generation, distribution and assignment models. Urban transit and inter-zonal modal split models. Consolidation of the three-step and four-step approaches. Role of the Bureau of Public Roads and US HUD. Land use - transportation studies and the development of land use models. Rise and fall of large scale models?

9 Chapter 3: Early developments in the UK Early 1960s: Transfer of the technology of forecasting from North America : zone-based to household class-based models : generalised costs and VoT expressions : Entropy maximising models introduced Empirical to analytic forms for location, mode and route splits. Much improved representation of public transit systems 1969 SELNEC Study: incorporates all major developments

10 Chapter 4: Developments in discrete choice models, Disaggregate models and methods based on rational choice. Major progress in forecasting approach based on specification, estimation, aggregation, and validation of discrete choice models. Wide application of multinomial logit (MNL) models particularly to modal choice and related problems. Early exploration of model structure (e.g., to combined choices). Increasing application of disaggregate behavioural approach in the US, UK and Netherlands. Contrasts between new and traditional approaches emphasized

11 Chapter 5: Further developments in discrete choice models, 1976 Major theoretical efforts in US and UK to address the problem of similarities between alternative choices Nested Logit and GEV family extensions established Traditional forecasting model reinvented in behavioural clothes. Anomalies found in forecasts! Numerical methods applied to address more complex choice processes. Micro-simulation methods used more widely. Stated preference methods developed and officially supported. Testing of disaggregate behavioural approach (USA and Netherlands). 1980s onwards. Major theoretical and empirical work on choice processes.

12 Chapter 6: Activity-based travel analysis and forecasting Late 1970s / early 1980s: broad critique of economic models as a basis for explanation and forecasting of travel behaviour. Empirical studies on complexity of choice processes and organisation of activities and travel in space and time Early 1990s: Towards a new generation of travel demand models. Tour-based and schedule-based approaches to activity-travel analysis. Econometric and computational process models approach operational status. Current status: Grand behavioural synthesis sought with marriage of choice-based and constraint-based approaches.

13 Chapter 7: Transportation network equilibrium models 1952: Wardrop s two criteria established. 1956: Beckmann-McGuire-Winsten network equilibrium model formulated. Exploration of combined model as an alternative to the traditional fourstep procedure. Classification of problems on basis of demand function (fixed/variable) and cost functions (symmetric / asymmetric). Solution algorithms proposed and convergence properties examined for various combined models. Rigorous solution procedures replaced fourstep heuristics. Feedback examined in detail. Challenges remain for multi-modal, multi-time period models with more realistic link cost functions.

14 Chapter 8: Tradition and Innovation in US practice Evolving requirements for urban travel forecasting (Clean Air Act Amendments, 1990). Responses of the MPOs limited and main tradition of practice unchallenged. Applications of tour-based and activity-based models. New travel forecasting initiatives TRANSIMS a work in progress (?). Major developments of applied land use - transportation models. Weaknesses of traditional urban goods movement models widely recognised: vehicle-based, commodity-based and agent-based approaches

15 Chapter 9: Tradition and innovation in UK practice Few comprehensive tour-based or activity-based models (e.g., PRISM). Incremental (pivot-point) models widely used and officially supported. Major studies on significant responses to policies and corresponding elasticities. Journey timing decision increasingly modelled. G/D/M/A no longer used as default G/M/D/A increasingly supported. Stated preference approaches widely adopted and officially supported. More network detail in space and time (e.g., SATURN, micro-simulation) Major advances in Integrated LU-T models

16 Chapter 10: Computing environment and travel forecasting software Our field and digital computing for civilian use grew up together. Development in hardware allowed innovations in models and software products. Developments of hardware for mainframe, mini- and micro-computers reviewed. Comparisons of computing speed, memory and cost. Travel forecasting software for mainframe, mini- and micro-computers examined. Travel forecasting software developers and products reviewed.

17 Chapter 11: Achievements, current challenges and future prospects Summarised progress in five stages. Challenges of behavioural modelling abound Still searching for internal consistency of complex models in practice. Assumptions underpinning forecasts and performance of models increasingly seen to be subject to challenge.

18 Conflicts and compromises between practice and research The role of the critic is vital for constructive innovation. Major challenges related to modeling new technological initiatives, innovative policies, etc. How will the field develop in an age of big data and new means of data collection?

19 Chapter 12: Conclusion How was progress achieved? - Progress through a series of paradigm changes? - Where do/did good ideas come from? Handing on the experience and learning in the process - How will future progress be judged? - Need to make our field more transparent to everyone, including interested citizens.

20 What happened next? We tried to consult widely on individual topics: Approximately 100 people approached for advice, and read selected passages 10 people read several chapters Draft of final manuscript submitted in June 2014 Published in UK in April 2015 Published in US and Rest of World in June 2015 Paperback edition will appear by mid We welcome comments / critiques on omissions / inaccuracies and biases dboyce@uic.edu williamsh@cardiff.ac.uk